How to Make Process Mining Stick? Attitudes for Successful Process Mining Initiatives

I recently read a fascinating research paper by Vinicius Stein Dani , Henrik Leopold , Jan Martijn van der Werf , Iris Beerepoot , and Hajo Reijers wittingly titled "From Loss of Interest to Denial: A Study on the Terminators of Process Mining Initiatives". The paper analyzes the reasons why process mining fails to "stick" in some organizations, and eventually fades out.

Based on a review of previous studies and over a dozen expert interviews, the researchers found that the following series of factors lead to termination of process mining initiatives.

First, after some initial insights, driven by the enthusiasm of a handful of internal stakeholders supported by a team of external consultants, comes the realization that data preparation and analysis take significant effort. My take here: This is particularly true when using tools that require SQL coding and thus round-trips between business teams and IT teams, hence the advantage of using a no-code process mining tool.

Next, as the external consultants move on, a lack of internal expertise and lack of focus leads to process mining insights coming slower and slower, leading to loss of management interest.

Along the way comes politics, with some stakeholders (e.g. line managers) denying the validity of findings that expose their deficiencies. A typical argument here is that "the data is incomplete", "this or that is not recorded", "we do this on time, we just don't record it right away", or (finger-pointing) "the problem is upstream, not in my part of the process".

But the ultimate killer is the lack of ROI, as process mining insights and improvement opportunities are not actioned. For example, opportunities for streamlining or automating some parts of a process, or preventing some types of defects require organizational changes and implementation efforts, which again get killed in politics.

Below are my recommendations on how to prevent the above killers of process mining initiatives.

First, it's important to keep in mind that process mining is an identifier, locator, and quantifier of business value. It's not a generator of value per se. It does not change the processes on its own. Ergo

• Recommendation 1. Ensure that process mining is done within the business team that will drive the changes, and not in e.g. a separate data science team. In other words, make sure the business teams can answer their own questions, validate their own hypotheses, and turn their own insights into action, i.e. put the "value identification" as close as possible to the "value actioning". The use of a no-code process mining tool can greatly help here.
• Recommendation 2. Avoid overprocessing in your process mining initiatives. Thinking iteratively. At each iteration, make sure you extract just enough data, do just enough data processing, build just enough analytics to identify value, move to process change as soon as value is located & quantified. Don't try to build the "perfect data processing pipeline" upfront. Don't try to launch large integration projects for data ingestion right away. Instead, iterate between identification phases and actioning phases.
• Recommendation 3. Do no limit process mining to analyzing the "as-is" situation. It's great to understand the past and the present, but equally important is to understand and quantify the benefit of implementing a process change. You are going to launch an effort to automate or streamline parts of your end-to-end process? Before you make a move, make sure you test the impact radius of this improvement effort. I have seen, so many times, operational excellence teams automating a task that appears to be the bottleneck, only to see an equally bad bottleneck emerging in another downstream task, leading to a zero net effect. Make sure you use your data not only to analyze the history, but also to simulate changes and quantify their impact, so that your action plans have predictable ROI. Data-driven simulation — simulation based on models discovered from data — is an excellent tool to rapidly check the benefit and impact of your improvement efforts before moving to implementation.
• Recommendation 4. Embed your process mining efforts within a data-driven business process management practice (or a Lean Six Sigma practice). Process mining is about methods & tools. It must be surrounded by strategic alignment and change management (governance, people, culture). People have to own the "continuous improvement cycle".

There are more recommendations to add here, but the above ones are a good starting point.

The paper makes for an insightful reading. The authors kindly provide a free copy of their paper here

• From Loss of Interest to Denial: A Study on the Terminators of Process Mining Initiatives - https://drive.google.com/file/d/1M-pQ1zWEC3hGBep0asCmkE3f4JVY8Qyf/view

Acks - Thanks to Maxim Vidgof for bubbling this paper in your Twitter feed.